Sql 防止循环和多个级联路径

我有以下SQL表：

create table dbo.Companies (
  Id int identity not null constraint primary key clustered (Id),   
);

create table dbo.Workers (
  Id int identity not null constraint primary key clustered (Id),  
  CompanyId int not null,
);

create table dbo.Evaluations (
  Id int identity not null constraint primary key clustered (Id),       
  CompanyId int not null,
  WorkerId int not null 
)

以及以下限制：

alter table dbo.Workers
add constraint FK_Workers_CompanyId foreign key (CompanyId) references dbo.Companies(Id) on delete cascade on update cascade;

alter table dbo.Evaluations
add constraint FK_Evaluations_CompanyId foreign key (CompanyId) references dbo.Companies(Id) on delete cascade on update cascade,
    constraint FK_Evaluations_WorkerId foreign key (WorkerId) references dbo.Workers(Id) on delete no action on update no action;

我正在使用实体框架，当我尝试插入一条记录时，我得到了错误：

Additional information: Introducing FOREIGN KEY constraint 'FK_dbo.Workers_dbo.Companies_CompanyId' on table 'Workers' may cause cycles or multiple cascade paths. Specify ON DELETE NO ACTION or ON UPDATE NO ACTION, or modify other FOREIGN KEY constraints.

我应该更改约束还是应该更改数据库设计

让我解释一下为什么我要这样设计我的数据库：

1-工人表中的公司ID表示工人当前工作的公司

2-评估中的公司ID表示进行评估时工人的工作地点。。。WorkerId显然表示被评估的工人

对我来说，这一切都是有道理的，但也许我错了……

更明确地说：

CREATE TABLE dbo.EventTypes (
      EID INT PRIMARY KEY IDENTITY(1,1)
    , EventDescription VARCHAR(40)
);
INSERT INTO dbo.EventTypes (EventDescription) VALUES ('Hired');
INSERT INTO dbo.EventTypes (EventDescription) VALUES ('Evaluated');

将使用这些数据并进行评估，以确定发生了什么，时间如下

CREATE TABLE dbo.Companies (
      CID INT PRIMARY KEY IDENTITY(1,1)
    /* ... */
);
CREATE TABLE dbo.Workers (
      WID INT PRIMARY KEY IDENTITY(1,1)
    /* ... */
);

CREATE TABLE dbo.History (
      HistID INT PRIMARY KEY IDENTITY(1,1)
    , TS DATETIME NOT NULL DEFAULT GETDATE()
    , CompanyID INT NOT NULL FOREIGN KEY REFERENCES dbo.Companies(CID) ON UPDATE CASCADE ON DELETE CASCADE
    , WorkerID INT NOT NULL FOREIGN KEY REFERENCES dbo.Workers(WID) ON UPDATE CASCADE ON DELETE CASCADE
    , EventTypeID INT NOT NULL FOREIGN KEY REFERENCES dbo.EventTypes(EID) ON UPDATE CASCADE ON DELETE CASCADE
    /* ...other generic history-traits... */
);
CREATE NONCLUSTERED INDEX Idx_History_TS ON dbo.History (TS) INCLUDE (CompanyID, WorkerID, EventTypeID) WITH (FILLFACTOR = 90);
CREATE NONCLUSTERED INDEX Idx_History_CompanyID ON dbo.History (CompanyID) WITH (FILLFACTOR = 90);
CREATE NONCLUSTERED INDEX Idx_History_WorkerID ON dbo.History (WorkerID) WITH (FILLFACTOR = 90);
CREATE NONCLUSTERED INDEX Idx_History_EventTypeID ON dbo.History (EventTypeID) WITH (FILLFACTOR = 90);

这将包含一系列事件，您可以根据自己的喜好对这些事件进行分类。例如，当关联的EventDescription='Evaluated'时，您的代码可以在此处查看：

CREATE TABLE dbo.Evaluations (
      EvalID INT NOT NULL PRIMARY KEY FOREIGN KEY REFERENCES dbo.History(HistID) ON UPDATE CASCADE ON DELETE CASCADE
    /* ...Eval columns... */
);

CREATE TABLE dbo.EvaluationItems (
      ID INT PRIMARY KEY IDENTITY(1,1)
    , EvalID INT NOT NULL FOREIGN KEY REFERENCES dbo.Evaluations (EvalID)
    /* ...item details... */
);
GO

这只是作为一个模型提供的，评估项目不必如此

其次，最近的雇主：

CREATE VIEW dbo.WorkersLastEmployer AS
    SELECT W.*, C.*
    FROM dbo.Workers W
        INNER JOIN (
            SELECT H.HistID, H.TS, H.CompanyID, H.WorkerID, MAX(TS)OVER(PARTITION BY H.WorkerID) AS LastHired
            FROM dbo.History H
                INNER JOIN dbo.EventTypes E ON E.EID = H.EventTypeID
            WHERE E.EventDescription = 'Hired'
        ) Hi ON Hi.WorkerID = W.WID
            AND Hi.TS = Hi.LastHired
        INNER JOIN dbo.Companies C on C.CID = Hi.CompanyID
GO

由于CompanyID已迁移到历史记录中，并且评估现在是一个直接子类，因此，公司Y的工人X评估视图：

CREATE VIEW dbo.WorkerEvalutations AS
    SELECT E.*, W.*, C.*
    FROM dbo.Evaluations E
        INNER JOIN dbo.History H on E.EvalID = H.HistID
        INNER JOIN dbo.Workers W on W.WID = H.WorkerID
        INNER JOIN dbo.Companies C on C.CID = H.CompanyID
GO

---

此结构运行时没有警告，并将为您提供一个更易于扩展的框架。

您的级联设置有两条通向dbo.companys的dbo.Evalutations的路径。我是否可以建议一种将员工与公司关系分离的设计，而是将这些关系建立在时间戳表中，并将评估作为同一时间戳表的子类，以便可以为评估或在新公司工作评估“EventType”列，在任何给定的评估期间，工作人员在哪里工作的视图是评估之前的第一个“更改”记录，按时间戳降序排列。CompanyID应该作为FK出现在时间戳表中。@JaazCole，对不起，我不太理解你关于时间戳的建议。你能提供一个代码示例吗。这很容易理解。非常感谢。